AI Broke Hiring: What It Means for SL Job Seekers
Both sides of hiring now run on AI: candidates mass-apply with it, recruiters screen with it. Here's what that arms race means if you're job-hunting from Sri Lanka.
AI hiring has quietly turned the job market into a machine talking to a machine. Candidates generate applications with AI, and recruiters filter them with AI, so the human moment that used to decide a hire keeps shrinking. Harvard Business Review made that case this month in AI Has Broken Hiring: Here's How to Fix It, and the headline lands because anyone who has applied for a job recently can feel it.
I want to set aside the "is AI good or bad" framing and ask the more useful question for someone applying from Colombo or Kandy: if both sides are automated, where does a real person actually win?
🔍 The arms race nobody asked for
The mechanics are simple, and that's the problem. A free language model can turn one CV into fifty tailored applications in an afternoon. So applicant volume per opening has exploded. Employers respond by leaning harder on automated screening to cope with the flood. Each side automates to keep up with the other, and the loop tightens.
Here's the asymmetry that matters for you:
| Side of the table | What AI gives them | What it costs them |
|---|---|---|
| Candidate | Volume — apply everywhere, fast | Signal — every application looks the same |
| Recruiter | Speed — filter thousands in seconds | Judgment — strong odd profiles get auto-rejected |
Key takeaway: When everyone uses the same tools to apply, the tools stop being an advantage. The edge moves back to whatever a model can't fake: a real portfolio, a referral, proof you shipped something.
The HBR piece argues the system is broken on both ends. I'd put it more bluntly: mass-applying with AI is now table stakes, not a strategy. If your whole plan is "let the model write it and spray," you're competing in the most crowded, lowest-signal lane there is.
🛠️ Beat the filter, then beat the human
Before a person reads your CV, software usually scores it. That layer is mechanical and gameable in the honest sense: it rewards clear structure, real keyword matches to the job description, and plain formatting it can parse. It punishes PDFs full of columns, images, and clever design.
A few things that reliably help at the machine stage:
- Mirror the job's own words. If the post says "React" and "REST APIs," use those exact terms where they're true. Don't stuff them, match them.
- Keep the layout boring. Single column, standard headings, no text trapped inside graphics.
- Quantify outcomes. "Cut page load from 4s to 1.2s" beats "improved performance."
- Name real tools and repos. A GitHub link is a claim a model can't manufacture for you.
You can pressure-test the machine stage for free. Our AI Resume ATS Checker scores how well a CV matches a specific job description, and the Resume Builder keeps the structure parser-friendly so you don't lose points on formatting alone.
Bottom line: Getting past the filter is necessary but not sufficient. It gets you to the human. The human is where you actually win or lose, so don't spend all your energy on the robot.
💡 The proof-of-work shift
If AI makes claims cheap, then claims become worthless and proof becomes everything. This is the part of the HBR argument I think will outlast the rest. When anyone can write "experienced full-stack engineer," the sentence stops meaning anything. What can't be auto-generated:
- A live project with a URL someone can click
- An open-source contribution with your name on the commit
- A short Loom or screen recording of you explaining a real decision you made
- A referral from a person who has actually seen your work
For a Sri Lankan engineer or student, this is genuinely good news. You don't need an expensive degree or a brand-name employer to produce proof. You need one or two things you actually built, hosted somewhere public, with the messy honest story of how you built them. That's more convincing to a hiring manager in 2026 than a flawless AI-polished CV, precisely because the CV is now assumed to be polished by a machine.
A practical move: build something small that solves a problem you personally have, ship it, and write up what broke. That single artifact does more than fifty generic applications.
🌐 What the broken system means for small teams
If you're on the hiring side of a small Sri Lankan team or startup, the same breakage hits you in reverse. You'll drown in AI-written applications that all read like the same competent stranger. Leaning on an automated filter feels efficient, but it quietly screens out the unconventional people who often turn out to be the best hires.
| Old hiring instinct | What I'd do now |
|---|---|
| Post widely, filter hard | Post narrow, ask for a small real task |
| Score the CV | Score the work sample |
| Trust the keyword match | Trust the 20-minute conversation |
A short paid task or a focused conversation cuts through the AI fog faster than any resume parser. It costs you time, but it surfaces exactly what the automated layer destroys: signal.
⚡ What this means for you
The takeaway isn't "stop using AI." I use it daily and so should you. The takeaway is that AI has flattened the easy moves, so the easy moves no longer differentiate anyone.
- If you're job-hunting: use AI to clear the filter, then win on proof a model can't fake. Check your CV against the actual job post, fix the structure, then go build the thing that makes the human say yes.
- If you're hiring: stop trusting the parser as a judge. Use it to sort, never to decide. Ask for a small piece of real work.
- For everyone: the value is moving back to the things that were always hard to fake. That's a fairer game than it sounds, and it favors people who actually do the work over people who just describe it well.
The system HBR calls broken is really just exposing what was always true. Now that the talking is automated, only the doing counts.
Original source
AI Has Broken Hiring